It allows using something other than $DISTRO_VERSION as the last
component of the repository name before the tag. This gives more
flexibility for organizing large collections of images e.g. for
different architectures / purposes.

See e.g. mesa/mesa!2722

Auto-picks yum vs dnf depending on availability, otherwise this is identical
to the fedora template.

For some reason, yum upgrade segfaults in a Centos 6 container (tested with
6.9 and 6.10) so this is disabled for now. Worth investigating for someone who
needs that container.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

I still can't get a reliable way to test that feature.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

There is a change in behavior regarding the skopeo command in the
after script. It now fails, meaning we are not tagging the image
properly.

At first, it seamed like a good idea to have a per-job tag, but
I can't find one single user of it, so I guess this is simply
unneeded anymore, and we should simply not tag with the Job ID.

Related: #3Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

Unclear why this works when run in the CI, but running any of these containers
with podman run --rm -it registry.freedesktop.org/.../ results in an error:

Error: chdir: No such file or directory: OCI runtime command not found error
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Alpine is interesting primarily because it uses musl, eudev and elogind which
can show some issues with erroneous include handling.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

or other branches might override the latest tag, which is not ideal
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

Arch doesn't seem to have official aarch64 images.

There is not much to add on the templates side. buildah will
automatically pull the correct arm64v8 image when running on the
aarch64 runner and with the correct image.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

If we want to support aarch64, we need an officially generated base
image, and archlinux is not part of https://hub.docker.com/u/arm64v8
AFAICT
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

See freedesktop/freedesktop#178 (comment 201897)Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

We want to allow to create and run VM with ci-templates. We are going to
need to duplicate the same behaviour, so better use an alias to avoid
copying the code.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

The current isolation method is working fine on local unprivileged
container, but doesn't seem to be working on unprivileged containers
on kubernetes with CRI-O.

Specifying the chroot isolation method allows to make sure we can run
our scripts in all situations.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

Because 29 will eventually die, and 30 is the new cool kid.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

We agreed on #1
that it should be MIT.

Fixes #1

We already disabled installation of recommended packages in
/etc/apt/apt.conf.

apt-get upgrade only upgrades installed packages, it doesn't install new
packages (to satisfy additional dependencies of newer versions of
installed packages) or remove installed packages (e.g. in favour of new
packages as part of a transition). dist-upgrade is needed to get the
latest version of all packages.

There is a lot of copy paste between the tests on the various templates.
And this is making the CI script quite long.

Split the various tests in separated files, so we can run a diff against
one another when adding a new template.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

Provide some basic usability and random notes.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>

This allows to run arbitrary command with parameters
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

It seems if 2 tags are getting written on the same registry, and if
those 2 tags are from the same base, then this confuses the registry.

The second attempt at rewriting it works, as now when the `push` is
started, the registry already has the blobs available, and it just
writes the manifest.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

Some CI scripts need to enable custom repositories, so better let them
decide what to do in the *_SCRIPT instead of forcing them.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

When running the Arch boostrap image under docker, for ubuntu 18.04 and
16.04 we get:

[root@0673f7a1e7b7 /]# buildah from ubuntu:18.04
Getting image source signatures
Copying blob 898c46f3b1a1 [======================================] 31.0MiB / 31.0MiB
Copying blob 63366dfa0a50 [======================================] 851b / 851b
Copying blob 041d4cd74a92 [======================================] 545b / 545b
Copying blob 6e1bee0f8701 [======================================] 162b / 162b
Copying config 94e814e2ef [======================================] 3.3KiB / 3.3KiB
Writing manifest to image destination
Storing signatures
The following failures happened while trying to pull image specified by "ubuntu:18.04" based on search registries in /etc/containers/registries.conf:
* "localhost/ubuntu:18.04": Error determining manifest MIME type for docker://localhost/ubuntu:18.04: pinging docker registry returned: Get https://localhost/v2/: dial tcp 127.0.0.1:443: connect: connection refused
* "docker.io/library/ubuntu:18.04": Error committing the finished image: error adding layer with blob "sha256:63366dfa0a5076458e37ebae948bc7823bab256ca27e09ab94d298e37df4c2a3": error creating layer with ID "f03e2da849b45acfcb4824e392d435bad4f8bc7520c225fd634644bc5980dac6": unknown file type for /var/lib/containers/storage/vfs/dir/762d8e1a60542b83df67c13ec0d75517e5104dee84d8aa7fe5401113f89854d9/dev/full
* "registry.fedoraproject.org/ubuntu:18.04": Error determining manifest MIME type for docker://registry.fedoraproject.org/ubuntu:18.04: Error reading manifest 18.04 in registry.fedoraproject.org/ubuntu: manifest unknown: manifest unknown
* "quay.io/ubuntu:18.04": Error determining manifest MIME type for docker://quay.io/ubuntu:18.04: Error reading manifest 18.04 in quay.io/ubuntu: error parsing HTTP 404 response body: invalid character '<' looking for beginning of value: "<!DOCTYPE HTML PUBLIC \"-//W3C//DTD HTML 3.2 Final//EN\">\n<title>404 Not Found</title>\n<h1>Not Found</h1>\n<p>The requested URL was not found on the server.  If you entered the URL manually please check your spelling and try again.</p>\n"
* "registry.access.redhat.com/ubuntu:18.04": Error determining manifest MIME type for docker://registry.access.redhat.com/ubuntu:18.04: Error reading manifest 18.04 in registry.access.redhat.com/ubuntu: name unknown: Repo not found
* "registry.centos.org/ubuntu:18.04": Error determining manifest MIME type for docker://registry.centos.org/ubuntu:18.04: Error reading manifest 18.04 in registry.centos.org/ubuntu: manifest unknown: manifest unknown

The arch based image works fine under podman, so not sure what is going
on.

Anyway, Fedora seems not affected by this bug, so switch to it.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

taken from https://gitlab.freedesktop.org/xorg/xserver/blob/c020769dbfb965740c8441d8242b738ef572a7c9/.gitlab-ci/DockerfileSigned-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

If the caller overrides the `before_script`, we will lose that information.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

`after_script` doesn't inherit the exported environment variables.
`before_script` was overriden by the subclasses, so pointless.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

It indeed hides useless data when everything goes well, but as soon as
there is an issue, we don't get the error message, so that's less than
ideal.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

Most of the code is similar between the templates.
We can not have a common file that we would include because multi-level
includes are not supported, but at least we can try making things
look similar so a diff should only show the actual packaging differences.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

https://gitlab.com/gitlab-com/support-forum/issues/4349Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

it's now a wayland project
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

the bootstrap image is now build from the official arch image, where
we just install buildah and podman. So there is no need to enforce
having a buildah:latest image in the local fork.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

We go from one image of 70MB to 700MB, but that's the price to pay
to not have troubles anymore.
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>

I still think it's best to not have a hidden file, but I am afraid that
this will lead to confusion when MR will not trigger the pipeline as
this requires a settings to be set on each clone.

So simply keep a symlink to the hidden file (this project is all about CI)
Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>